Exploring the Impact of Human Behaviour in Common Collision Scenarios Based on Digital Twin Simulations

In urban environments, vehicles share the road space with other road user types, like vulnerable road users (VRU), i.e., pedestrians and bicyclists. Nowadays, preventing accidents with VRUs requires inferring a variety of information about a VRU from the perspective of the vehicle. This prevention is challenging for at least two reasons: First, pedestrians have higher degrees of freedom of movement and can spontaneously alter their movement dynamics compared to vehicles. Second, in the urban environment, “non-line-of-sight” (NLOS) situations are common. In NLOS situations VRUs are not within the field of view of the driver and car-based sensor systems. Consequently, the VRU may not be identified in time, thus leading to a collision.

The avoidability of collisions between VRUs and vehicles in urban NLOS scenarios highly depends on the behavior of both the VRU and the vehicles, respectively. Thus, as part of the research project KoSi (funded by the BMDV), we integrate human behavior models in the 3D simulation environment \({\text {Tronis}}^\circledR \). Considering individual human behaviors in this virtual prototyping and validation environment allows to asses the performance of collision avoidance systems more realistically in contrast to generalized human behavior models. The significance of considering different human behaviors is highlighted by results of simulations for the Euro NCAP NLOS scenario with different behavior models for the driver and a pedestrian, respectively.